Corporate Earnings: It’s not just a question of quantity

September 30, 2007 Introduction outline the findings of our own research and illustrate They say beauty is only skin-deep. Investors would be how we have incorporated analysis into wise to remember this aphorism when choosing stocks our investment process. since a company’s stated earnings, though eagerly watched by the market, are only ink-deep. To truly un- Background derstand a firm’s financial situation, it is important to The success Smith Group has achieved over the years consider the quality of a firm’s earnings, not just the is attributable to our ability to control portfolio risk and quantity. invest in companies whose earnings grow faster than the market expects. Because reported earnings are so The Basics... important to our process, and because managements have so many tools available to them to Understanding the quality of earnings is important to forecasting the quantity and credibility of fu- manipulate earnings, it is important that we assess the ture earnings. credibility of earnings for every company we invest in our clients’ portfolios. Level of , operating margins, turnover and exclusions are significant indicators of the degree of earnings quality. The foundations of our earnings quality model rest on a large array of research generated both externally and Earnings quality measures differ by industry, making internally. We have studied numerous academic papers customization of rankings models necessary. from some of the leading minds in the field. The out- Companies with high earnings quality have a come of this review was a confirmation of the primary higher likelihood of reporting positive earnings measures that we already use, but with some enhance- surprises in the future. ments that give us improved insight into a company’s earnings quality at an industry level. This enhancement There is evidence that company management teams added a great deal of strength to the model’s perform- overly fixate on meeting short-term earnings growth ance by eliminating or overweighting certain measures expectations. When these earnings targets are at risk, according to their specific relevance to an industry. many companies manipulate accounting rules to create For example, banks clearly generate earnings in a dif- the appearance of continued growth. Unfortunately, ferent way than other businesses, so customizing the aggressive accounting practices used to pump up earn- measurements to determine the quality of earnings for ings today will have to be reversed at some point in the a bank is important. This type of industry-level cus- future. This opens up the possibility that an accurate tomization can make the difference between enhancing assessment of earnings quality can be used to achieve performance and detracting from it. superior investment returns. From an investment analysis perspective, high quality earnings should be: Academic work has revealed certain measures that can offer insight into the quality and sustainability of a firm’s earnings. In addition, Smith Group has con- • Truthful – Earnings growth derived from account- ducted its own research in this area, building upon ing manipulation does not represent a company’s true prior work. We have verified, refined and expanded fundamentals and is of little use for investors. the findings, and used these results to enhance our pro- prietary earnings quality ranking model. • Sustainable – Only earnings from the ongoing activi- ties of a company’s operations can be expected to recur This paper will summarize relevant past research re- in future periods. Earnings that come from an atypical lated to the subject of earnings quality. It will then activity, such as the one time sale of a large asset, do not form a good basis for prediction.

Academic Papers sions, and earnings that result from a high level of exclu- Recent academic literature has suggested metrics that can sions are of questionable quality. be used to measure earnings quality. To summarize, the quality of earnings can be successfully

measured by such variables as: In Richard Sloan’s 1996 paper “Do Stock Prices Fully • Reflect Information in Accruals and Flows About Accruals Future Earnings?”, he finds two effective measures of • Profit margin earnings quality: accruals and cash flow. Accruals are the • Asset turnover difference between reported earnings and cash flow; they often result from and which have been • Expense exclusions booked but for which cash has not yet changed hands (i.e. accounts receivable and payable). The paper con- These insights bolstered our existing beliefs and pointed cludes that companies with high use of accruals are less us toward possible enhancements that could be made to likely to show persistence in reported earnings growth. our earnings quality model. On the other hand, companies with relatively low use of accruals are more likely to have high quality and sustain- The Smith Group Earnings Quality Model able earnings. Sloan’s work identified mispricing associ- We began the project by testing existing components of ated with this variable. our earnings quality model as well as additional measures presented in academia and other research. We measured Additional insight into earnings quality can be found in them based on their intuitive ability to explain earnings Mark Soliman’s 2002 paper “Using Industry-Adjusted quality as well as their historical ability to predict future Dupont Analysis to Predict Future Profitability .” This return. All measures fit within our traditional approach to analysis looks at two components of ROA (return on fundamental analysis. Tests were conducted at the sector ): profit margin and asset turnover and investigates and industry level. their ability to predict future returns. The result is an improved ranking model tailored individu- The results show that companies with high profit margins ally to the nature of each company’s business. In all cases and high asset turnover relative to their industry are likely we let theory guide and validate our efforts. Blind data to have positive earnings growth in the future due to a mining turns up relationships that are coincidental and do company’s competitive edge. Soliman concludes that im- not hold up outside the sample in which they are found. proving asset turnover is a particularly persistent factor We only accepted as valid those relationships which both and a good predictor of repeat earnings. exhibited strong statistical significance and made sense from a theoretical point of view. These measures provide In a 2003 paper Jeffrey Doyle, Russell Lundholm and us with a window into the future sustainability of earnings Mark Soliman looked at the predictive value of expense which is not reflected by current stock prices. exclusions from pro forma earnings to assess the quality of these “non-standardized” earnings reports.

While the practice has abated somewhat over the past few years, many firms prefer to announce pro forma earnings instead of GAAP (generally accepted accounting principals) earnings. Pro forma earnings often exclude certain expenses from earnings calculations creating the appearance of higher profitability and more attractive valuation. The stated reasoning given by management is that adjusted earnings give investors a better view of the core, ongoing activities of the enterprise. More pessimis- tic observers claim the intention is merely to inflate earn- ings. The chart above illustrates the consistency of market re- turn relative to our investment universe to stocks ranked The paper finds that excluded expenses do, in fact, have a in the top 20 percent according to our earnings quality significant negative relationship to future earnings. Thus, model. Many stock selection factors perform well in ei- firms do not appear to be justified in their use of exclu- ther growth or value cycles(such as 2000 – 2002) but rarely

Page 2

in both. However, our EQ model performs remarkably depth fundamental analysis. Qualitatively, portfolio man- well in distinctly different types of market environments. agers use the components of the earnings quality model This analysis is useful in determining when the factor to identify areas of concern that may warrant further fun- outperforms or underperforms and establishing how damental analysis. well the model might interact with other factors we use in our investment process. Combination with Other Ranking Models

The How We Use Earnings Quality in Our In- Our model works well on its own and when incorporated into our existing stock selection process. The earnings vestment Process quality score can independently generate alpha, helping Our earnings quality model helps us at both ends of our identify companies that will generate positive earnings investment process – identifying companies that outper- surprises and sustainable growth. Additionally, the meas- form as well as underperform the market. ure enhances our ability to avoid companies that are more likely to restate earnings, something almost always Because companies with poor earnings quality rankings associated with significant negative returns. have a higher likelihood of underperforming the market, we eliminate them from consideration through quantita- The true test of any fundamental factor we consider is tive screening. whether it contributes to enhancing the performance pat- tern we deliver to our clients. Our “Growth Prediction” On the positive side, the ranking model is very good at models focus on companies that will grow faster than the identifying companies that are more likely to show earn- market expects. Screening on this factor in addition to ings growth persistence and to report positive earnings high earnings quality has a compounding effect, provid- surprises. Because of these benefits, our portfolios are ing a significant boost to the model’s ability to deliver made up of companies with much better earnings quality alpha. than the benchmark. The bar chart below measures the historical return From an individual company level, earnings quality spreads associated with each model individually and com- analysis provides us a window into the direction of a bined. Companies with the highest Growth Prediction company’s financial position and future earnings and Earnings Quality rankings outperform those with the growth. We prefer companies that are decreasing their worst combined rankings by over 30% per year. use of accruals, improving the efficiency in which they operate their businesses and generating increased levels of free cash flow.

Deterioration of these trends results in a lower model score and is a “red flag” that warrants further analysis. An existing holding that experiences a material decline in earnings quality will be sold. Our sell discipline is focused on eliminating companies that have a high likeli- hood of reporting a negative and our research shows this occurs more often in companies with low and declining earnings quality.

Our earnings quality model is used both quantitatively Further evidence of the value derived from combining and qualitatively. We use quantitative screening to elimi- the two factors is supported by the chart on the following nate companies from consideration for long positions page. The most attractive companies based on the that increase risk in the portfolio. For selection of short “Growth Prediction” model beat quarterly earnings esti- positions, the opposite is true. Low earnings quality mates 72% of the time whereas companies that are companies make great shorting candidates for our ranked attractively by both models beat quarterly earnings 130/30 portfolios (see related white paper). Screening estimates 82% of the time. This is a significant increase for companies with both attractive earnings quality and in surprise prediction given that the average stock beats improving earnings growth outlook help refine our final earnings estimates nearly 60% of the time. candidate list that is used as a starting place for more in-

Page 3

Disclosures Founded in 1995, Smith Asset Management Group, L.P. (“Smith Group”) is a registered investment advisor that specializes in investment management services.

Earnings Quality/Earnings Surprise: According to many academic studies as well as our own research, earnings quality and earnings surprise have had a positive relation- ship to relative performance in certain time periods for certain companies. However, this does not mean that this relationship exists for all time periods and for all companies. In the past, periods coinciding with an in- verse relationship between earnings surprise and earnings Combination with Other Ranking Models quality, and relative performance have typically been peri- Earnings quality analysis has long been part of the re- ods in which corporate earnings are not the focus of in- search process at Smith Group. It gives us an additional vestors’ attention. Additionally, companies which have tool to work with in selecting stocks and building supe- had a chronic negative relationship between earnings sur- rior portfolios for our clients. It allows us to put a prise and relative performance are typically those compa- firm’s earnings data “under the microscope” to gain in- nies whose earnings are not product-driven, such as com- sight into future earnings growth. This is necessary in a modity companies. There is no assurance that the his- world where the earnings picture painted by manage- toric positive relationship between earnings surprise, ment can be distorted through the intricacies of modern earnings quality, and relative performance will exist in the accounting procedures. Intelligent investing requires future. Nor is there any assurance that the historic ability analysis below the surface to the true beauty, or lack of Smith Group to forecast a high rate of positive earn- thereof, that lies beneath. You can rest assured that at ings surprise and high earnings quality companies will Smith Group, we look at both the quantity and quality of exist in the future. earnings. The pro-forma performance results presented in this study do not represent the results of actual trading using client or firm assets, but were achieved by means of the retroactive application of a computer model that was de- signed with the benefit of hindsight. Pro-forma perform- ance results such as those shown in this presentation have significant limitations, and should not be relied 100 Crescent Court, Suite 1150 upon as indicative of future performance. Pro-forma Dallas, Texas 75201 Phone: 214-880-4600 performance is presented before estimated exchange fees, Our website address is: www.smithasset.com commissions and trade execution costs. During the time periods presented in this study, Smith Group, L.P. was not managing assets in a dedicated earnings quality fo- cused investment strategy.

This material is for your own personal information, and we are not soliciting any action based upon it. The mate- rial is based upon information we consider reliable, but we do not represent that it is accurate or complete and it should not be relied upon as such. Opinions included in this material are as of September 30, 2007 and are subject to change without prior notice.

Past performance is not indicative of future results. As with any investment vehicle, there is always a potential for profit as well as the possibility of loss. Actual results may differ from composite returns, depending on ac- count size, investment guidelines and/or restrictions, in- ception date and other factors. Nothing contained in this presentation should be construed as a recommendation to buy or sell a security or economic sector. Page 3